We reinterpret the semi-supervised data classification problem using an auction dynamics framework (inspired by real life auctions) in which elements of the data set make bids to the class of their choice. This leads to a novel forward and reverse auction method for data classification that readily incorporates volume/class size constraints into an accurate and efficient algorithm requiring remarkably little training/labeled data. We prove that the algorithm is unconditionally stable.

This presentation is part of Minisymposium “MS4 - Graph Techniques for Image Processing (2 parts)”
organized by: Yifei Lou (University of Texas at Dallas) , Jing Qin (Montana State University) .

Authors:

Ekaterina Rapinchuk (Michigan State University)

Keywords:

machine learning